Diabetes Care e1

Sodium–Glucose Cotransporter Devin Abrahami,1,2 Antonios Douros,1,3,4 Hui Yin,1 2 Inhibitors and the Risk of Fractures Oriana H.Y. Yu,1,5 and Among Patients With Type 2 Diabetes Laurent Azoulay1,2,6 https://doi.org/10.2337/dc19-0849

The association between sodium–glucose (alternate metformin indication) at the initiation before 2013; i.e., prevalent cotransporter 2 (SGLT2) inhibitors and the time of the first-ever prescription of an use); and 4) other treatments. Current risk of fractures is controversial. In the antidiabetic drug. Using this base and new use DPP-4 inhibitors served as Canagliflozin Cardiovascular Assessment cohort, we then assembled a study the reference group. For all exposure Study (CANVAS) Program, canagliflozin cohort of individuals who initiated a new categories, exposed person-time was de- was associated with a significant in- antidiabetic drug class as of 2013 (year fined by the prescription duration plus a creased risk of fractures compared with the first SGLT2 inhibitor was intro- 90-day grace period to allow for a poten- placebo (hazard ratio [HR] 1.26, 95% CI duced in the U.K.). Cohort entry was tial latent exposure effect on the out- 1.04–1.52) (1). Possible mechanisms may defined by the date of this new pre- come. Time-dependent Cox proportional involve elevated serum phosphate levels scription. We excluded individuals hazards models were used to calculate or reductions in bone mineral density (2). with a diagnosis of Paget disease, HRs, adjusted for 45 covariates (Table 1), To date, three recent observational stud- osteomalacia, or hyperparathyroid- including smoking status, BMI, medica- ies investigated this association, but ism, rare risk factors associated with tions, and duration of treated diabetes. these did not observe an increased risk fracture incidence, at any time before This study was approved by the CPRD of fractures (3–5). However, these cohort entry. Patients were monitored Independent Scientific Advisory Commit- studies had some limitations, including from cohort entry until the first of an tee (protocol 18_221) and by the Jewish residual confounding, a limited outcome incident diagnosis of fracture, death from General Hospital Research Ethics Board. definition, or restriction to a single SGLT2 any cause, an incident diagnosis of Paget This study included 73,178 patients, inhibitor. disease, osteomalacia, or hyperparathy- with 9,454 SGLT2 inhibitor users (1,288 To address these limitations, we roidism, end of registration with the canagliflozin, 5,539 dapagliflozin, 2,133 conducted a population-based cohort general practice, or end of the study empagliflozin, and 494 users of .1 study using the U.K. Clinical Practice period (31 March 2018). SGLT2 inhibitor) and 18,410 DPP-4 in- fi fi Research Datalink (CPRD). We rst We de ned exposure using a time- hibitor users, monitored for a median e-LETTERS assembled a base cohort of all individ- varying definition, where each person- duration of 1.9 years (patient character- uals, at least 40 years old, newly treated day of follow-up was classified into one istics on file). There were 1,973 fracture with antidiabetic drugs between 1 Jan- of the following four mutually exclusive events, corresponding to an incidence – uary 1988 and 31 December 2017. We categories: 1) current use of SGLT2 in- rate of 12.88 (95% CI 12.32–13.46) per OBSERVATIONS excluded individuals with ,1yearof hibitors; 2) current use of dipeptidyl 1,000 person-years. medical history, those initially pre- peptidase 4 (DPP-4) inhibitors (with ini- The results of this study are presented scribed insulin in monotherapy, and tiation after 2013; i.e., new use); 3) in Table 1. Compared with use of DPP-4 women with polycystic ovary syndrome current use of DPP-4 inhibitors (with inhibitors, use of SGLT2 inhibitors was not

1Centre for Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, Montreal, Canada 2Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Canada 3Department of Medicine, McGill University, Montreal, Canada 4Institute of Clinical Pharmacology and Toxicology, Charite´ – Universitatsmedizin¨ Berlin, corporate member of Freie Universitat¨ Berlin, Humboldt-Universitat¨ zu Berlin, and Berlin Institute of Health, Berlin, Germany 5Division of Endocrinology, Jewish General Hospital, Montreal, Canada 6Gerald Bronfman Department of Oncology, McGill University, Montreal, Canada Corresponding author: Laurent Azoulay, [email protected] Received 29 April 2019 and accepted 23 June 2019 © 2019 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license. Diabetes Care Publish Ahead of Print, published online July 11, 2019 e2 SGLT2 Inhibitors and Fracture Risk Diabetes Care

Table 1—Crude and adjusted HRs for the association between the use of SGLT2 inhibitors and the risk of fractures HR Analysis Events Person-years Incidence rate (95% CI)* Crude Adjusted† Primary analysis‡ DPP-4 inhibitors 403 27,210 14.8 (13.4–16.3) 1.00 [Reference] 1.00 [Reference] SGLT2 inhibitors 138 12,500 11.0 (9.3–13.0) 0.75 0.97 (0.79–1.19) Type of SGLT2 inhibitor§ Canagliflozin only 10 1,764 5.7 (2.7–10.4) 0.38 0.47 (0.25–0.88) Dapagliflozin only 98 8,624 11.4 (9.2–13.8) 0.77 1.01 (0.81–1.27) Empagliflozin only 30 2,031 14.8 (10.0–21.1) 1.00 1.29 (0.88–1.88) Type of fracture| Upper limb 57 12,500 4.6 (3.5–5.9) 0.85 1.00 (0.73–1.38) Lower limb 46 12,500 3.7 (2.7–4.9) 0.68 0.93 (0.66–1.32) Vertebral 11 12,500 0.9 (0.4–1.6) 1.23 1.76 (0.81–3.82) Other 33 12,500 2.6 (1.8–3.7) 0.57 0.89 (0.60–1.33) Effect measure modification¶ No fracture history 129 12,170 10.6 (8.8–12.6) 0.74 0.95 (0.78–1.18) History of fracture 9 330 27.3 (12.5–51.8) 0.95 1.18 (0.51–2.69) No history of osteoporosis 130 12,180 10.7 (8.9–12.7) 0.80 0.97 (0.78–1.20) History of osteoporosis 8 320 25.0 (10.8–49.3) 0.67 0.84 (0.39–1.81) ,65 years old 83 8,957 9.3 (7.4–11.5) 0.94 0.91 (0.69–1.22) 65–74 years old 45 3,063 14.7 (10.7–19.7) 1.13 1.11 (0.78–1.59) $75 years old 10 480 20.8 (10.0–38.3) 0.77 0.84 (0.44–1.61) Female 68 4,845 14.0 (10.9–17.8) 0.63 0.88 (0.67–1.17) Male 70 7,654 9.1 (7.1–11.6) 0.93 1.10 (0.82–1.47) Sensitivity analyses Competing risk 138 12,500 11.0 (9.3–13.0) 0.76 0.98 (0.80–1.20) High-dimensional disease risk score# 138 12,500 11.0 (9.3–13.0) 0.75 0.97 (0.80–1.18) Marginal structural model** 137 153,736 0.9 (0.7–1.1) 0.73 0.95 (0.73–1.24) Alternate reference group†† 138 12,500 11.0 (9.3–13.0) 0.97 1.10 (0.82–1.48) *Per 1,000 person-years. †Adjusted for age, sex, year of cohort entry, alcohol-related disorders (including alcohol dependency, alcoholic cirrhosis of the liver, alcoholic hepatitis, and hepatic failure), smoking status, BMI, inflammatory bowel disease, rheumatoid arthritis, hemoglobin A1c, duration of treated diabetes, macrovascular complications of diabetes (including stroke, myocardial infarction, and peripheral vascular disease), microvascular complications of diabetes (including nephropathy, neuropathy, and retinopathy), cancer, fracture history, osteoporosis, and use of bisphosphonates, use of antidiabetic drugs in the year before cohort entry, use of calcium and vitamin D, number of unique nonantidiabetic drugs in the year before cohort entry, use of benzodiazepines, sedative-hypnotics, opioids, selective serotonin reuptake inhibitors, tricyclic antidepressants, anti-Parkinson disease drugs, centrally acting antihypertensive drugs, nitrates, nonnitrate antianginal agents, thiazide and thiazide-like diuretics, inhaled glucocorticoids, oral glucocorticoids, proton pump inhibitors, H2 receptor antagonists, anticonvulsants, atypical antipsychotics, typical antipsychotics, and loop diuretics in the year before cohort entry. Variables with missing data were imputed using multiple imputation, using results from five imputations. ‡The median duration of DPP-4 inhibitor use was 366 days, and the median duration of SGLT2 inhibitor use was 343.5 days. Current use of DPP-4 inhibitors (with initiationbefore 2013)anduse of othertreatmentswereconsideredin the modelbut notpresentedinthe table.Thesegroups generated67 eventsfrom 4,357 person-years and 1,365 events from 109,113 person-years, respectively. §Drug switchers (individuals who used more than one type of SGLT2 inhibitor) were considered in a separate category. This group generated 0 events from 81 person-years. |DPP-4 inhibitor users (reference) generated 148, 146, 19, and 128 upper-limb, lower-limb, vertebral, and other fractures, respectively, in 27,210 person-years of follow-up. ¶DPP-4 inhibitor users without a fracture history generated 382 events in 26,481 person-years, and those with a history of fracture generated 21 events in 729 person-years. DPP-4 inhibitorusers without a history of osteoporosis generated337 eventsin 25,471person-years, and those with a history of osteoporosisgenerated 66 events in 1,739 person-years. DPP-4 inhibitor users aged ,65 years old generated 124 events in 12,567 person-years, users aged 65–74 generated 108 events in 8,318 person-years, and users aged $75 generated 171 events in 6,325 person-years. Female DPP-4 inhibitor users generated 242 events in 10,879 person-years, and male DPP-4 inhibitor users generated 161 events in 16,331 person-years. #Adjusted for high-dimensional disease risk score, which included all variables listed above and 200 variables empirically selected by the high-dimensional disease risk score algorithm. **This model considered person-months of follow-up and generated marginal HRs. DPP-4 inhibitor users generated 406 events in 336,321 person-months. ††Use of glucagon-like peptide-1 receptor agonists was considered as the reference group, which generated 65 events during 5,675 person-years of follow-up.

associated with fracture risk (HR 0.97, 95% or sex. Finally, the results remained con- In summary, the results of this large CI 0.79–1.19). When stratifying by SGLT2 sistent in several sensitivity analyses inves- population-based cohort study indicate molecule, use of canagliflozin was asso- tigating death as a competing risk (using that use of SGLT2 inhibitors is not asso- ciated with a decreased risk of fractures the Fine and Gray method), when con- ciated with an increased risk of fractures (HR 0.47, 95% CI 0.25–0.88), but this trolling for confounding using the high- compared with use of DPP-4 inhibitors. finding was based on few events (n 5 dimensional disease risk score, when Although our stratified analysis by SGLT2 10). There were no differences in risk using a marginal structural model with molecule revealed a protective effect for according to fracture type, except for inverse probability of treatment and canagliflozin, this finding should be a numerically elevated HR with vertebral censoring weighting, or when compar- interpreted with caution given that it fractures (HR 1.76, 95% CI 0.81–3.82). ing to glucagon-like peptide-1 receptor was based on few exposed events. Over- Similarly, the association was not modified agonists, as was done in a previous all, our real-world study provides fur- by history of fracture, osteoporosis, age, study (5). ther reassurance on the safety of this care.diabetesjournals.org Abrahami and Associates e3

new class of drugs on bone health Author Contributions. D.A. wrote the manu- Canagliflozin and cardiovascular and renal among patients with type 2 diabetes. script. D.A., A.D., H.Y., O.H.Y.Y., and L.A. critically events in type 2 diabetes. N Engl J Med 2017; revised the manuscript. All authors conceived and 377:644–657 designed the study, analyzed and interpreted the 2. Faillie JL. Pharmacological aspects of the Funding. This research was funded by a Foundation data, approved the final version of the manuscript, safety of gliflozins. Pharmacol Res 2017;118: Scheme grant from the Canadian Institutes of Health and are accountable for its accuracy. D.A., H.Y., 71–81 Research (FDN-333744). D.A. is the recipient of a and L.A. did the statistical analyses. L.A. acquired 3. Schmedt N, Andersohn F, Walker J, Garbe E. Vanier Canada Graduate Scholarship from the Ca- the data and supervised the study. L.A. is the Sodium-glucose co-transporter-2 inhibitors and nadian Institutes of Health Research and recipient guarantor of this work and, as such, had full the risk of fractures of the upper or lower limbs of a training grant from the Canadian Institutes of access to all the data in the study and takes in patients with type 2 diabetes: a nested case- Health Research Drug Safety and Effectiveness Train- responsibility for the integrity of the data and control study. Diabetes Obes Metab 2019;21: ingProgram.O.H.Y.Y.holds aChercheur-Boursier the accuracy of the data analysis. 52–60 Clinicien Junior 1 award from the Fonds de Re- Prior Presentation. Partsofthisstudywere 4. Ueda P, Svanstrom¨ H, Melbye M, et al. cherche du Quebec´ – Sante.´ L.A. holds a Chercheur- presented as a poster presentation at the Sodium glucose cotransporter 2 inhibitors and Boursier Senior award from the Fonds de Recherche American Diabetes Association's 79th Scien- risk of serious adverse events: nationwide du Quebec´ – Sante´ and is the recipient of a William tific Sessions, San Francisco, CA, 7–11 June register based cohort study. BMJ 2018;363: Dawson Scholar Award from McGill University. 2019. k4365 DualityofInterest. L.A. receivedconsulting fees 5. Fralick M, Kim SC, Schneeweiss S, Kim D, from Janssen for work unrelated to this study. References Redelmeier DA, Patorno E. Fracture risk after No other potential conflicts of interest relevant 1. Neal B, Perkovic V, Mahaffey KW, et al.; initiation of use of canagliflozin: a cohort study. to this article were reported. CANVAS Program Collaborative Group. Ann Intern Med. 2019;170:155–163